Skip to main content
Unlisted page
This page is unlisted. Search engines will not index it, and only users having a direct link can access it.

Databricks Feature Lookup Client release notes

This page lists releases of the Databricks feature lookup client, available on PyPI. This library is used by served models to look up feature values from the Databricks Feature Store. It is not intended to be installed by users.

databricks-feature-lookup 1.19.0​

  • Improve serving performance for rolling-window (sawtooth) aggregations.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.18.0​

  • Return null instead of an empty list for rolling-window (sawtooth) FIRST_N and LAST_N aggregations over an empty window, consistent with offline training sets.
  • Support rolling-window (sawtooth) APPROX_COUNT_DISTINCT and APPROX_PERCENTILE aggregations. These aggregations require a model logged with Python 3.9 through 3.13.
  • Support serving models and feature specs with a large number of feature tables, including tables spread across multiple Lakebase databases.
  • Preserve exact types of complex feature values served from Lakebase, including map keys and FLOAT precision. A nested value that doesn't match its saved type now raises an error that names the feature and field, instead of being converted.
  • Accept omitted complex RequestSource inputs when the model signature marks them optional.
  • Return a clear error when the online store for a feature lookup was deleted or is unreachable, instead of a generic model-input error.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.17.0​

  • Return 0 for count-based Feature View aggregations when no matching row is found in the online store, consistent with offline materialization.
  • Preserve the original types of complex feature values stored in Lakebase before model inference.
  • Improve Lakebase lookup resilience by retrying transient errors and reusing connections after transient conflicts.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.16.0​

  • Support additional rolling-window (sawtooth) aggregations: FIRST/LAST and variance/standard deviation.
  • Improve Lakebase lookup latency with over 10 feature tables by using pipelined queries.
  • Support online feature view names over 63 characters by leveraging shortened column names in Lakebase.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.15.0​

  • Support DECIMAL type features in SQL and Lakebase lookups.
  • Support serving rolling-window (sawtooth) and adjacent-day feature lookups.
  • Serve Unity Catalog UDF-backed RequestSource features via built-in feature functions.
  • Improve feature serving latency (parallel Lakebase fetches, reduced GC pauses).
  • Remove legacy Online Tables support.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.14.0​

  • Support STRUCT type features in lookup client.
  • Support OpenTelemetry (beta) tracing for Feature Serving endpoints.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.13.0​

  • Support RequestSource features for model serving.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.12.0​

  • Handle error in secrets rotation.
  • Reduce frequency of Postgres connections recycling.
  • Support tracing with Lakebase query.

databricks-feature-lookup 1.1.0​

  • Support performance metrics instrumentation.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.0.1​

  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 1.0.0​

  • Improve performance to reduce lookup latency.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.9.0​

  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.8.1​

  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.8.0​

  • Security enhancements for connections to Databricks Online Stores.
  • Support for DAG of data and transformation nodes in feature pipelining.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.7.0​

  • Instance profile authentication supported for lookup with Feature Store models.

databricks-feature-lookup 0.6.0​

  • Support for Databricks Online Stores added.
  • Improvements configured for DynamoDB and CosmosDB retry policies.

databricks-feature-lookup 0.5.1​

  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.5.0​

  • Support for Feature Functions at lookup time added.

databricks-feature-lookup 0.4.1​

  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.4.0​

  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.3.0​

  • Performance improvements to reduce latency of batch online lookup from DynamoDB.

databricks-feature-lookup 0.2.0​

  • Reduce number of database connections used when connecting to Amazon Aurora and Amazon RDS MySQL.
  • Performance improvements to reduce latency of online lookup when overriding features.
  • Miscellaneous bug fixes and improvements.

databricks-feature-lookup 0.1.2​

  • Fix compatibility issues with protobuf >= 4.21.0.

databricks-feature-lookup 0.1.1​

  • Look up feature values from Amazon DynamoDB.
  • Look up the latest feature values from time series feature tables published in window mode.
  • Miscellaneous bug fixes and improvements.